U.S. Department of Housing and Urban Development Office of Policy Development and Research
All Other Things Being Equal:
A Paired Testing Study of Mortgage Lending Institutions
Final Report
April 2002
All Other Things Being Equal: A Paired Testing Study of Mortgage Lending Institutions
Final Report
Prepared for:
U.S. Department of Housing and Urban Development Office of Fair Housing and Equal Opportunity
By:
Margery Austin Turner Fred Freiberg
Erin Godfrey Carla Herbig Diane K. Levy Robin R. Smith
The Urban Institute Washington, DC 20037
April 2002
ACKNOWLEDGEMENTS
This project could not have been completed without the commitment and hard work of testers, testing coordinators, and local testing organizations in Los Angeles, Chicago, and New Orleans. The authors therefore extend our appreciation and thanks to the directors and staff of Progressive Management Resources, Inc., the Leadership Council for Metropolitan Open Communities, and the Greater New Orleans Fair Housing Action Center.
Stephen L. Ross, of the University of Connecticut, served as an adviser to this project, assisting in the design of testing protocols and tester assignments, and reviewing draft
materials, results, and reports. His expertise in the areas of housing finance, paired testing, and statistical methods has been invaluable.
In addition, a panel of experienced fair housing professionals provided advice and guidance on testing protocols and procedures. We thank Tracey Goodson Barrett of the Housing and Civil Enforcement Section of the Civil Rights Division of the U.S. Department of Justice, Mary Davis of the Leadership Council for Metropolitan Open Communities, Jonathan Jones and Heidi Olguin of Progressive Management Resources, Inc., Rick Rhey of the Southern Arizona Housing Center, and Carla Wertheim of the Metropolitan Milwaukee Fair Housing Council.
Finally, staff of the Department of Housing and Urban Development, particularly Dale Rhines and Sharon Bower from the Office of Fair Housing and Equal Opportunity, provided guidance and oversight throughout the project’s design and implementation. In addition, Todd Richardson from the Office of Policy Development and Research assisted in the final review process.
Despite the generous contributions from these individuals and organizations, any errors and omissions that may remain in this report are, of course, our own.
U . S . D E P A R T M E N T O F H O U S I N G A N D U R B A N D E V E L O P M E N T W AS H I N G T O N , D . C . 2 0 4 1 0 - 0 0 0 1
T H E S E C R E T A R Y
Foreword to “All Other Things Being Equal"
HUD Secretary Martinez
A new home buying report, "All Other Things Being Equal," conducted in Los Angeles and Chicago shows that most people of color do not face discriminatory treatment when they inquire about loan products. Using testers, the report uncovered that both minority and white testers received equal treatment the majority of the time. However, the findings show that African-American and Hispanic homebuyers face a statistically significant risk of receiving less favorable treatment than comparable whites when they ask mortgage-lending institutions about financing options.
The study focused on lending discrimination in the earliest stage of the homebuying process – when a prospective homebuyer first begins to apply for a mortgage loan. If an individual is discouraged from applying for a loan at all, or is given discriminatory loan terms and conditions, the entire home buying process is fatally compromised.
Under contract with the U.S. Department of Housing and Urban Development, the Urban Institute used a paired testing technique to determine if minority home loan applicants received the same treatment and information at the pre-application stage of the process as white
applicants. Hispanic and African-American testers were matched with white testers who had roughly equivalent financial backgrounds.
“All Other Things Being Equal" underscores that there are still those who would
contradict not only Fair Housing laws, but the principles of freedom and opportunity we treasure as Americans. The Department will use these findings to help assure that American homebuyers and the mortgage lending industry are aware of their rights and responsibilities under the Fair Housing Act.
Mel Martinez
TABLE OF CONTENTS
EXECUTIVE SUMMARY ... i
CHAPTER 1: BACKGROUND AND APPROACH ... 1
The Homeownership Testing Project ... 2
Overview of the HTP Methodology... 4
Organization of the Report ... 6
CHAPTER 2: THE HTP PAIRED TESTING METHODOLOGY... 8
Sample of Lending Institutions ... 8
Basic Test Structure ... 12
Organizational Structure... 12
Test Authorizations... 13
Tester Assignments - Characteristics and Instructions ... 13
Testing Protocols... 16
Replicating HTP Lender Testing Protocols ... 18
CHAPTER 3: PATTERNS OF UNEQUAL TREATMENT IN LOS ANGELES AND CHICAGO ... 20
Statistical Methods for Paired Testing Analysis... 21
Question 1: Did testers receive the information they requested?... 24
Question 2: How much were testers told they could afford? ... 26
Question 3: How many loan products were testers told about? ... 27
Question 4: How much “coaching” did testers receive? ... 29
Question 5: Did testers receive follow-up contact? ... 31
Question 6: Were testers encouraged to consider FHA? ... 34
Summary of Statistically Significant Findings ... 36
CHAPTER 4: SUMMARY AND CONCLUSIONS... 37
Pre-Test Stage ... 37
Pilot Stage ... 39
Future Paired Testing Research ... 40
ANNEX A: ASSIGNED FINANCIAL CHARACTERISTICS ANNEX B: TESTING REPORT FORM
ANNEX C: TERMS AND CONDITIONS
EXECUTIVE SUMMARY
When first-time homebuyers begin shopping for a house, they need to learn about mortgages for which they can qualify and about house prices they can afford. This information can be provided by a variety of different sources, including mortgage lending institutions, real estate agents, and mortgage brokers. But if potential homebuyers cannot obtain full and fair access to information about mortgage financing, they may give up on their pursuit of
homeownership, their housing search may be restricted, or they may be unable to negotiate the most favorable loan terms. Thus, pre-application inquiries about mortgage financing options represent a critical phase in the homebuying process.
The U.S. Department of Housing and Urban Development contracted with the Urban Institute to rigorously assess the effectiveness of paired testing for determining whether minority homebuyers receive the same treatment and information as whites at the pre-application phase of the mortgage lending process, and to produce rigorous measures of the incidence of unequal treatment in two metropolitan areas. The mortgage lending process consists of a complex series of stages, including advertising and outreach by lending institutions, responses to pre- application inquiries from potential borrowers, approval or denial of loan applications and determination of loan terms and conditions, and finally, loan administration. Discrimination may occur at any of these stages and may take different forms at different stages. 1
In a paired test, two individuals―one white and one minority―pose as homebuyers and inquire about the availability and terms for home mortgage loans. Because the two members of a tester team present themselves as equally qualified borrowers in every respect except their race or ethnicity, systematic differences in the treatment they receive provide direct evidence of adverse treatment discrimination. Paired testing has been used widely to detect and measure discrimination by rental and sales agents, but only a few, relatively small-scale investigative studies have been applied to mortgage lending.
This study consisted of two major stages―a pre-test stage and a pilot stage. The pre- test stage was used to experiment with a fairly wide variety of paired testing scenarios, and to assess the feasibility of testing several different sources of mortgage financing information.
More specifically, we conducted tests of mortgage lending institutions, mortgage brokers, real estate agents, new construction sales agents, and mobile home sales agents. This pre-test experience informed the design of the pilot stage, in which conducted approximately 250 paired tests of a representative sample of mortgage lending institutions in Los Angeles, California and
1 For more information on existing evidence about discrimination at other stages in the mortgage lending process, see Margery Austin Turner and Felicity Skidmore.1999. Mortgage Lending Discrimination: A Review of Existing Evidence. Washington, D.C.: The Urban Institute.
Chicago, Illinois, using a standardized set of protocols in order to yield statistically rigorous measures of discrimination against African Americans and Hispanics in these two metropolitan housing markets.
Paired Testing Can Be an Effective Tool for Research and Enforcement
In the first stage of this study, a total of 78 tests were conducted in two markets―
Orange County, California and New Orleans, Louisiana. Because the format of these tests, and the information sources they targeted were so varied, they cannot be used to produce statistical measures of the extent of discrimination. Nevertheless, the pre-tests do provide important insights about the forms that mortgage lending discrimination may take and about methods for conducting effective testing.
The intensive pre-testing effort found that rigorous paired testing is feasible―though difficult―for all the information sources we considered. The pilot stage focused on mortgage lending institutions, because of the central role they play in the sector. But future research and enforcement efforts can and should explore other information sources, including mortgage brokers and referrals by real estate agents. More research is also needed to fully understand how homebuyers search for mortgage financing, including racial and ethnic differences in information sources and search strategies.
The pre-test effort also found that one of the most serious forms of discrimination that can be discerned by paired testing at the pre-application stage is differential estimates of home price and total loan amount. These estimates potentially play a major role in determining where people search for housing and whether they decide they can afford to become homebuyers.
Therefore, the pilot phase of the Homeownership Testing Project was explicitly designed to measure this form of differential treatment rigorously.
Finally, pre-testing highlighted the complexity and difficulty of mortgage lending testing.
This type of testing demands more from both testers and testing organizations than routine rental testing, and great care must be taken with tester training, supervision, and record
keeping, as well as with the actual conduct of the tests themselves in order to achieve credible test results. Not all local fair housing organizations necessarily have the capacity to effectively conduct lending tests.
African Americans Experience Unequal Treatment from Lending Institutions in Los Angeles and Chicago
The pilot phase of the Homeownership Testing Project focused on a single, clearly defined scenario: A first-time homebuyer visits a mortgage lending institution requesting help in figuring out a price range for housing, a loan amount for which he or she might qualify, and loan products that might be suitable, to answer the basic research question: Do minorities receive
the same treatment and information as whites at the pre-application phase? All the tests conducted in the pilot phase of the Homeownership Testing Project consisted of individuals visiting offices of mortgage lending institutions (including both conventional and sub-prime lenders). Testers posed as first-time homebuyers, making a general uninformed request for information about how much house they could afford and what loan products might be available to them. All of the testers were assigned financial profiles that qualified them for products targeted to borrowers with A- credit in their respective housing markets. Their income and asset levels qualified them to purchase a median-priced house in their metropolitan area, but they were asset constrained and had minor problems with their credit history. The two members of each tester pair were assigned virtually identical financial and household characteristics, with the minority partner always slightly better qualified than the white.
In both Chicago and Los Angeles we conducted tests for a representative sample mortgage lending institutions in the metropolitan area that report under the Home Mortgage Disclosure Act (HMDA),2accept at least 90 mortgage loan applications per year, and had offices in the region that a first-time homebuyer could realistically find and visit. Lending institutions with very large application volumes not only had a high probability of selection, but could appear in the sample more than once. This sampling strategy allows us to make statements about the incidence of differential treatment by large lending institutions in Chicago and Los Angeles that are directly accessible to first-time homebuyers.3
The pilot test results show that in both Los Angeles and Chicago, African American and Hispanic homebuyers face a significant risk of receiving less favorable treatment than
comparable whites when they visit mortgage lending institutions to inquire about financing options. In the majority of cases, minorities and whites received equal treatment, or when differences occurred, they were equally likely to favor the minority as the white.4 But in both
2 HMDA requires all independent mortgage companies and mortgage lenders owned by depository institutions that make at least 100 home purchase and/or refinancing loans in a given year to report on the demographic and locational characteristics of all applications and loans.
3 Levels and patterns of discrimination may be different for smaller lending institutions or when inquiries are made by telephone or internet rather than in person.
4 It is important to recognize that even when we do not observe a statistically significant pattern of systematically unequal treatment, discrimination may have occurred in individual cases. For some treatment variables, differences in treatment occurred quite frequently but favored minorities just as often as whites. These differences may result from random variations in the behavior of loan officers, but they may also include cases of discrimination on the basis of race or ethnicity.
metropolitan areas, paired testing reveals statistically significant patterns of unequal treatment that systematically favor whites.5
Unequal treatment takes different forms in the two metropolitan areas and for the two minority groups.
In Los Angeles --
• Blacks were offered less coaching than comparable white homebuyers, and were more likely to be encouraged to consider an FHA loan.
• Hispanics were denied basic information about loan amount and house price, told about fewer products, and received less follow-up compared to Anglo homebuyers.
In Chicago –
• Blacks were denied basic information about loan amount and house price, told about fewer products, offered less coaching, and received less follow-up than comparable white homebuyers.
• Hispanics were quoted lower loan amounts or house prices, told about fewer products, and offered less coaching than comparable Anglo homebuyers.
These patterns of unequal treatment occurred regardless of whether the two members of a tester pair met with the same loan officer or with different loan officers.
Adverse Treatment of African American and Hispanic Homebuyers Face a
Los Angeles Chicago
Treatment Categories
Blacks Hispanics Blacks Hispanics
Information requested U U
Loan amount and house price U
Number of products U U U
Coaching U U U
Follow-up contact U U
FHA encouraged U
U Unequal treatment favors whites or Anglos over blacks or Hispanics at a 90 percent level of statistical significance or higher
No statistically significant differences in treatment
5 Results are reported as statistically significant when the difference between the rate of white/Anglo-favored treatment and the rate of minority-favored treatment is significant at a 90 percent confidence level or higher.
The frequency of unfavorable treatment varies considerably from one category to another. For some categories, equal treatment occurred in the vast majority of cases, but when differences occurred, the white tester was dramatically more likely to be favored than his or her minority partner.
For other treatment categories, differences in treatment occurred much more often, but again were substantially more likely to favor the white than the minority tester. The series of charts below illustrate the levels of
unfavorable treatment for each treatment category where statistically significant differences were observed.
Despite the variations in the forms that differential treatment takes, these findings demonstrate that African American and Hispanic homebuyers in both Los Angeles and Chicago face a significant risk of unequal treatment when they visit main- stream mortgage lending institutions to make
pre-application inquiries. Discriminatory treatment at this early stage in the mortgage lending process has the potential to discourage some minorities from continuing their housing search, to limit their search to lower cost homes than they could actually afford, and to prevent them from choosing the most favorable loan products.
Therefore, we conclude that paired testing at the pre-application stage of the mortgage lending process is feasible and effective for both enforcement and research purposes.
Additional testing should be conducted, including systematic studies that focus on other sources of information about mortgage products and on homeseekers who are less well-qualified as borrowers. However, because of the complexity of this kind of testing and the differences between metro area results, we do not recommend a large-scale national study of discrimination at the pre-application stage of the mortgage lending process. Instead, systematic studies should be conducted on a site-by-site basis, with ample time and resources for effective training, test coordination, and quality control.
Two female testers, one white and one black, visited the same Los Angeles area lender two days apart and met with the same loan officer. The testers told the loan officer that they were first-time homebuyers and needed assistance in figuring out a home price range and a loan amount for which they might qualify. The loan officer requested and obtained detailed information on household income, debts, and assets from both testers and asked about their respective credit situations. He then estimated that the white tester would qualify for a $332,500 loan to purchase a $350,000 home and, but estimated that the black tester would qualify for a $237,500 loan to purchase a $245,000 home. The loan officer told the white tester that a seller would likely pay some of the closing costs, but no mention was made about seller assistance to the black tester. The loan officer also told the white tester that it was a good idea to have a home inspection conducted prior to purchase, while the loan officer did not mention anything about the value of a home inspection to the black tester. The loan officer provided a complete loan application package to the white tester, but not to the black tester.
CHAPTER 1: BACKGROUND AND APPROACH
More than three decades after the passage of the Fair Housing Act (Title VIII of the 1968 Civil Rights Act), African American and Hispanic homebuyers still do not enjoy equal access to home ownership. Widespread evidence indicates that minority homebuyers are less likely than whites to obtain mortgage loans and, if they are successful, receive less favorable loan amounts and terms. However, questions remain about the extent to which discrimination is the cause of these unequal outcomes, and about the forms that discrimination may take. A recent review of existing social science evidence concluded that minority homebuyers in the United States do face discrimination from mortgage lending institutions.1 But serious gaps remain in what we know and more rigorous information is needed to refine and target enforcement strategies, to enable lending institutions to monitor their own performance, and to design remedies to reduce discrimination in home mortgage lending.
Paired testing provides a uniquely powerful tool for investigating both the incidence and the forms of adverse treatment based on race or ethnicity. In a paired test, two individuals―one white and one minority―pose as homebuyers and inquire about the availability and terms for home mortgage loans.2 Because the two members of a tester team present themselves as equally qualified borrowers in every respect except their race or ethnicity, systematic differences in the treatment they receive provide direct evidence of adverse treatment. Paired testing has been used widely to detect and measure adverse treatment by rental and sales agents, but only a few, relatively small-scale investigative studies―primarily by the National Fair Housing
Alliance (NFHA)―have been applied to mortgage lending. During the early 1990s, NFHA conducted tests in seven cities (Atlanta, Chicago, Dallas, Denver, Detroit, Oakland, and Richmond). Testers posed as first-time homebuyers and refinancers inquiring about financing terms and conditions at the pre-application stage. A reanalysis of these testing data by the Urban Institute concluded that differential treatment occurred at significant levels in at least some cities. Minorities were less likely to receive information about loan products, received less time and information from loan officers, and were quoted higher interest rates in most of the cities where tests were conducted.
1 Mortgage Lending Discrimination: A Review of Existing Evidence. 1999. Margery A. Turner and Felicity Skidmore, Eds. The Urban Institute.
2 For more information on paired testing and its role in both measurement and enforcement, see A National Report Card on Discrimination in America: The Role of Testing. 1999. Michael Fix and Margery Austin Turner, Eds.
The Urban Institute.
The Homeownership Testing Project
Building upon the experience and findings of previous testing efforts, the Department of Housing and Urban Development (HUD) contracted with the Urban Institute to further refine and apply methods for conducting paired testing at the pre-application stage of the mortgage lending process. The Homeownership Testing Project (HTP) was designed to meet two basic
objectives. The first is to develop and refine a set of testing methodologies that will provide guidance to others who might be interested in testing for lending discrimination. The second objective of the Homeownership Testing Project is to produce reliable and representative indicators about the nature and extent of differential treatment by home mortgage lending institutions in two major metropolitan areas. Because the NFHA testing was designed primarily for enforcement purposes, its findings do not provide statistically rigorous measures of the incidence of differential treatment across lending institutions.3 Thus, this project combines research and enforcement goals, and seeks to advance both knowledge about the incidence of mortgage lending discrimination and tools for conducting rigorous testing at the pre-application stage.
Because of the complexity of the mortgage application process and the challenges it presents for paired testing, we divided this effort into two basic stages―a pre-test stage and a pilot stage. The pre-test stage was used to experiment with a fairly wide variety of paired testing scenarios, and to assess the feasibility of testing several different sources of mortgage financing information. A total of 78 tests were conducted in two markets―Orange County, California and New Orleans, Louisiana. Although these pre-tests cannot be used to produce statistical measures of the extent of adverse treatment, they do provide important insights about the forms that mortgage lending discrimination may take and about methods for conducting effective testing.
The pre-test stage explored the feasibility of conducting paired testing for six different information sources, or “portals,” reflecting the range of possible sources of mortgage
information used by homebuyers:
• Conventional mortgage lenders serve large numbers of homebuyers directly. In addition, homebuyers who access other portals, such as real estate agents, often end up meeting with a lender at some point in the financing process.
3 In particular, lenders were selected for inclusion in the NFHA study based on evidence suggesting that they might be discriminating against minority customers. Thus, the findings are not necessarily representative of treatment by all mortgage lenders. See Robin Smith and Michelle DeLair (1999). “New Evidence from Lender Testing: Discrimination at the Pre-Application Stage,” in Margery Austin Turner and Felicity Skidmore (eds), Mortgage Lending Discrimination: A Review of the Exisitng Evidence. Washington, D.C.: The Urban Institute.
• Sub-prime mortgage lenders are, according to some researchers and policymakers, the fastest growing segment of the mortgage industry. This growth is due, in part, to the increasing numbers of homebuyers unable to secure a conventional loan due to blemished credit histories or other financial characteristics that will not meet the underwriting criteria of conventional lenders.
• Mortgage brokers provide the same services as mortgage lending institutions, up to the point of actually originating a loan. In some markets, and for some groups of borrowers, they may be an important source of information on loan products. However, it is difficult to develop a complete and reliable list of active mortgage in a metropolitan area from which to draw a representative sample. Moreover, since many brokers do business primarily through referrals, future testing will need to experiment with credible scenarios for testers to contact brokers.
• Real estate agents who have their own financing arm can provide customers with financing information, while other agents may refer homebuyers to outside mortgage brokers or lenders for additional information or for the actual loan. They are important gateways to the mortgage industry and influence which lenders many buyers contact. Conventional sales testing of real estate agents can be expanded to assess information provided about mortgage financing.
• New construction sales offices and mobile home dealers both provide customers with information about financing terms and conditions. These portals serve limited, but important, segments of the housing market. Testing for pre-application lending
discrimination by sales agents and mobile home dealers can and should be incorporated into more comprehensive tests for discrimination by these actors. In other words, tests designed to detect differences in unit availability, costs, amenities, and location can also incorporate indicators of differences in financing amounts, terms, and conditions.
The pre-tests were also used to experiment with different testing scenarios. Some testers indicated that they already had a particular house in mind while others were trying to find out how much they could potentially afford. Some testers posed as well-qualified borrowers, while others were assigned marginal qualifications. Some testers visited lenders in pairs, posing as husband and wife inquiring together, while others conducted one-person visits, but indicated that they were married.
This intensive exploratory effort concluded that rigorous paired testing is feasible for all the portals we considered. Future research and enforcement efforts can and should explore other portals, including mortgage brokers and referrals by real estate agents. The pre-test effort also found that one of the most serious forms of adverse treatment that can be discerned by paired testing at the pre-application stage is differential estimates of home price and total loan amount. These estimates potentially play a major role in determining where people search for
housing and whether they decide they can afford to become homebuyers. Finally, pre-testing highlighted the complexity and difficulty of mortgage lending testing. This type of testing
demands more from both testers and testing organizations than routine rental testing, and great care must be taken with tester training, supervision, and record keeping, as well as with the actual conduct of the tests themselves in order to achieve credible test results.
Lessons from the pre-test phase have been incorporated into a package of tools for enforcement testing, produced as part of the Homeownership Testing Project.4 In addition, the pre-test experience informed the design of the project’s pilot stage, in which we conducted approximately 250 paired tests of a representative sample of mortgage lending institutions in Los Angeles, California and Chicago, Illinois. These tests followed a single, standardized set of protocols in order to yield statistically rigorous measures of adverse treatment against African Americans and Hispanics in the two metropolitan housing markets.
Overview of the HTP Methodology
Mortgage lending is a tremendously complex process. It can involve multiple information providers, including real estate agents, brokers, and different types of lending institutions. Many lenders offer a wide range of loan products, with varying interest rates, points, downpayment requirements, repayment terms, and monthly costs. And a number of different customer characteristics, including income, assets and debts, and credit history, help determine what products are offered. This complexity makes it challenging to design and conduct paired tests that reliably measure the incidence of differential treatment. It would be impossible to design a single study protocol to capture all the possible variations, even at the pre-application stage.
The pilot phase of the Homeownership Testing Project focused on a single, clearly defined scenario: A first-time homebuyer visits a mortgage lending institution requesting help in figuring out a price range for housing, a loan amount for which he or she might qualify, and loan products that might be suitable, to answer the basic research question: Do minorities receive the same treatment and information as whites at the pre-application phase? This section outlines the key elements of a rigorous paired testing methodology to measure the incidence and severity of adverse treatment by mortgage lending institutions in two very different housing markets.
All the tests conducted in the pilot phase of the Homeownership Testing Project consisted of individuals visiting offices of mortgage lending institutions (including both
conventional and subprime lenders). We focused on mortgage lending institutions for two basic reasons. First, mortgage lenders are an important source of information about loan products,
4 See Fred Freiberg and Carla Herbig. 2001. Guide to Enforcement Tools for Fair Lending Testing.
Washington, D.C.: The Urban Institute.
terms, and affordability. Homebuyer surveys indicate that a majority of both minority and white homebuyers speak to a mortgage lender at some point during the process of purchasing a home. Even homebuyers who begin their search for information elsewhere are likely to end up working with a mortgage lending institution at some point in the process. And second, mortgage lending institutions can be identified and sampled with reasonable precision, in order to produce statistically rigorous measures of adverse treatment. Our pre-test experience indicated that designing valid samples of sufficient size for other information portals would be difficult.
All of the testers in the pilot phase posed as first-time homebuyers, making a general uninformed request for information about how much house they could afford and what loan products might be available to them. They visited lenders’ offices after calling to make an appointment. Although most real homebuyers gather information by telephone, almost one third report making personal visits to lenders, and our pre-test experience indicated that in-person visits would be both feasible and credible. Moreover, we experimented in the pre-test phase with a number of variations on this basic scenario and found that testers do not need to identify a house, neighborhood, price range, or loan product in order to be provided with information by mortgage lenders. Pre-testing also indicated that individuals posing as one member of a married couple were treated as serious customers.
The pilot testing effort was implemented in two metropolitan areas, Chicago and Los Angeles. In both of these sites, sufficient numbers of tests were conducted to produce
statistically rigorous estimates of differential treatment for blacks (compared to whites) and for Hispanics (compared to Anglos). Four basic criteria guided the selection of these sites for the study’s pilot phase:
• Strong local testing capacity. Although at least 100 fair housing groups nationwide conduct paired testing, relatively few are likely to have the capacity to conduct large numbers of lender tests to the rigorous standards of this project. Therefore, a principal criterion for selecting sites was the availability of a local testing organization with strong capacity. Chicago’s Leadership Council for Metropolitan Open Communities and the Law Firm of Sharon Lybeck Hartmann in Los Angeles are well-established testing organizations, with experienced testers and supervisors and the institutional capacity to take on this challenging project.
• Geographic diversity. Although two sites obviously cannot be statistically representative of metro areas nationwide, we wanted to select pilot sites that are sufficiently different to reflect the diversity of metropolitan housing markets. Chicago and Los Angeles
represent different geographic regions and have very different housing market conditions.
• Size and minority population. We limited our site selection to large metropolitan areas with substantial black and Hispanic populations. Our pre-test experience reinforced
earlier assumptions that the anonymity of large urban areas helps shield a large-scale testing effort from detection, and that a substantial minority population lends credibility to minority testers. Chicago and Los Angeles have sufficient numbers of both African Americans and Hispanics to permit testing for adverse treatment against both groups.
• Number of lenders. Given the volume of testing required to yield statistically rigorous measures of differential treatment, it is critical that sites have a large number of mortgage lenders with local offices. Otherwise, the project might risk detection by conducting too many tests of the same lending institutions.
In both Chicago and Los Angeles we conducted tests for a representative sample of all the mortgage lending institutions in the metropolitan area that accepted a significant number of mortgage loan applications and had offices in the region that a first-time homebuyer could realistically find and visit. Lending institutions with very large application volumes not only had a high probability of selection, but could appear in the sample more than once. This sampling strategy allows us to make statements about the incidence of differential treatment by primary lending institutions in Chicago and Los Angeles that are directly accessible to first-time
homebuyers.
The treatment that potential homebuyers receive from lending institutions is highly sensitive to the financial profiles they present. Therefore, the two members of each tester pair were assigned virtually identical characteristics for all our pilot tests. In addition to the potential problems posed by differences between testing partners, the variation in profiles across tester pairs could influence the results of the pilot phase. Because the number of tests that could reasonably be conducted in each site was relatively small (by statistical standards), we decided to keep the tester profiles homogeneous. In other words, we limited the extent to which results across tests might vary as a result of differences in income levels, assets and debts, or credit histories. By reducing the potential variation between tests, we increased the statistical
reliability of findings from each site’s sample. All of the testers were assigned financial profiles that qualified them for products targeted to borrowers with A- credit in their respective housing markets. Their income and asset levels qualified them to purchase a median-priced house in their metropolitan area, but they were asset constrained and had minor problems with their credit history.
Organization of the Report
The remainder of this report consists of three chapters. Chapter 2 more fully explains the sampling procedures and paired testing methodology implemented in the pilot phase of the Homeownership Testing Project, building upon the experience gained in the pre-test phase.
Chapter 3 presents the results of the pilot phase, summarizing measures of differential treatment for each of the racial/ethnic tracks in our two pilot sites: black/white tests in Los
Angeles; black/white tests in Chicago; Hispanic/Anglo tests in Los Angeles; and Hispanic/Anglo tests in Chicago. Finally, Chapter 4 concludes the report with a summary of key findings from both the pre-test phase and the pilot phase of the Homeownership Testing Project and their implications for future research and enforcement efforts.
CHAPTER 2: THE HTP PAIRED TESTING METHODOLOGY
Conducting successful pre-application mortgage lending tests requires a highly structured and controlled process that includes careful preparation prior to the test visits, conscientious adherence to assignments during test visits, and thorough documentation of the test experience after test visits are completed. This chapter first describes the methodology used to select a representative sample of lending institutions for testing in each site. We then present the test structure that was employed during the pilot phase, the assignment of tester characteristics and instructions to testers, and the testing protocols used to conduct pre- application mortgage lending. We conclude by offering some general observations about replicating the HTP testing protocols used during the pilot phase and issues related to
organizational capacity that are required to conduct effective pre-application mortgage lending testing.
Sample of Lending Institutions
The Homeownership Testing Project was designed to (1) evaluate lender testing
methodologies in two pre-test sites and (2) estimate the incidence of discrimination by mortgage lenders in two pilot sites. The distinct objectives of each of these two components imply
different guidelines for selecting the lenders to be tested.
The pre-test focused both on tester characteristics and on the entry points at which testers gain access to information on mortgage products. At this stage, our goals were to implement and refine testing forms and protocols to find out whether different testing scenarios were effective and what types of differential treatment they could detect. To operationalize this phase, lenders were selected on a case-by-case basis, to meet the needs of the specific test scenario being explored. Lists of different types of lenders were gleaned from Home Mortgage Disclosure Act data, telephone books, local newspapers, and home guides. In the pre-test, there was no need to select a strictly random or representative sample of lenders.
Measuring the frequency of discrimination in the pilot phase required a more systematic selection of lenders so that statistically reliable measures of the incidence of differential
treatment across a large number of tests could be produced. For these findings to be generalized to particular types of local lenders (or a particular segment of a local lending market), a sampling frame was required that fully reflected local mortgage lending activity and provided an equal chance for each lender to be selected for testing. By using a rigorous sampling frame, test results can answer questions about the incidence (or likelihood) of
discrimination for a given population in a particular situation. While the pre-test phase accessed information about mortgage financing through a variety of portals such as mortgage lenders, mobile home dealers, real estate agents, and new home builders, the pilot phase focused
lender selection could not occur on a case-by-case basis and instead, was driven by an overarching sampling frame of local mortgage lending institutions from which all entities to be tested were selected.
No sampling frame is perfect; some compromises and caveats are inevitable. Ideally, however, our goal was to construct an exhaustive (and mutually exclusive) list of all institutions making mortgage loans in a metropolitan market from which all lenders to be tested would be selected. To generate the closest approximation of such a list for the pilot phase, we used information reported under the Home Mortgage Disclosure Act (HMDA) in 1998.
HMDA data are available for all mortgage companies owned by depository institutions and independent mortgage companies that make at least 100 home purchase and/or refinance loans in a given year. These data include information on the number of home mortgage loans made by a lender, characteristics of applicants, and locations of prospective properties. HMDA data for a specific locality list all lenders (within the proscriptions above) making loans for properties in that locality.
In reviewing information from HMDA reports, it is apparent that many lenders are listed who receive very few applications for mortgage loans. This raised concerns about the feasibility of testing. Two people visiting such an office within a short period of time with a similar request could be very unusual. To better ensure that testing activity would not prompt suspicion, we selected for the sampling frame only those institutions that received at least 90 mortgage loan applications in a given year. For Los Angeles, this meant 344 of 515 institutions were dropped from the sampling frame. However, the remaining 171 institutions conducted the vast majority of the application activity captured in HMDA (97 percent). Similarly, in Chicago we included in the sampling frame 198 of 793 institutions covering 94 percent of application activity.
In order for a tester to make an in-person visit to a lending institution, the lender must have a local office. Moreover, that office and its services must be accessible to the average consumer without requiring membership or ties to a specific organization. For the sampling frame to reflect these needs, we deleted from the list of all HMDA reporters those institutions (1) without local offices and (2) unavailable to the average consumer (such as credit unions).
Once the list of lenders was pared down to those institutions receiving at least 90
applications a year that had an office available for a visit from an average consumer, institutions were called to verify both the presence of a local office and the ability of a person to receive mortgage information at that office. An institution without an office, that was out of business, or that did not provide information to potential first-time borrowers, was deemed inaccessible and unavailable for testing. In Los Angeles, this brought the number of lenders listed in HMDA who were available for testing from 171 to 67. These 67 lenders conducted roughly half of the mortgage loan application activity reported in HMDA. In Chicago, 106 lending institutions were
available for testing, reflecting 62 percent of the market’s HMDA-reported application activity (Exhibit 2-1).
Exhibit 2-1: Sampling Frame
# of Institutions Application Volume % of HMDA Activity
LA Chicago LA Chicago LA Chicago
Full HMDA List 515 793 160,518 167,231 100 100
90+ Applications 171 198 155,730 157,002 97 94
Accessible 67* 106** 89,788 103,017 56 62
*In addition to deleting credit unions and lenders without local offices, this category included 35 institutions that were no longer in business, 7 wholesalers, and 4 that did not serve first-time homebuyers.
**Again, in addition to deleting credit unions and lenders without local offices, this category included 2 institutions that were no longer in business, 15 wholesalers, 6 that did not serve first-time homebuyers, and 5 that would only pre-qualify applicants over the phone.
The sampling frame used in the homeownership testing project can be generalized to HMDA reporting lenders who receive at least 90 applications a year and are accessible to the average consumer. While this sampling frame covers a significant portion of the mortgage activity in a given locality, it does not reflect entities that do not report to HMDA. Because independent mortgage companies making fewer than 100 homes loans do not have to report for HMDA purposes, the sample excludes smaller independent mortgage companies. It is unclear how many institutions and how much activity this misses in Los Angeles and Chicago. In addition, mortgage brokers connect many prospective borrowers to potential lenders, but because they do not actually make loans, they are not listed in HMDA data.
Research testing seeks to simulate as closely as possible the typical experience of a mortgage credit seeker in the marketplace. Therefore, the ideal sampling frame is one which duplicates the market entry points used by the average person when searching for mortgage credit. The more closely the sampling frame resembles the credit seekers’ experience, the more accurately it measures the treatment of typical potential borrowers.
Unfortunately, there is little in-depth research on how people search for credit including potentially different search practices of racial and ethnic groups. While most research testing efforts have used lenders as the entry point for mortgage credit, other housing professionals such as real estate agents may provide significant gate keeper functions in the actual mortgage market.
Moreover, different segments of the mortgage industry may position themselves to serve niche markets of potential borrowers. To the extent that targeted marketing results in more
aggressive campaigns to individuals or communities of different racial groups, this could affect the type of institutions used by borrowers.
To generate a sample of lenders for testing, institutions were sampled with replacement from the restricted population of mortgage lending institutions, with the probability of selection proportionate to volume of loan applications. This means that an institution with a high volume of loan applications was more likely to be selected than one with a low volume of applications.
Moreover, large institutions could be drawn into the sample more than once, so that multiple tests were conducted for these lenders. This approach resulted in a sample that includes both large and small lenders but gives large lenders (which serve a larger share of the market) a greater weight in the ultimate test results.6 From the 67 institutions in the Los Angeles sampling frame, 35 were selected for black/white testing and 34 for Hispanic/Anglo testing. In both categories, the lenders selected accounted for half of the application activity captured in HMDA data. In Chicago, the lenders selected for black/white testing cover 47 percent of the application activity, while those tested for treatment of Hispanics and Anglos cover 51 percent.
Exhibit 2-2: Final Sample of Lending Institutions
# of Institutions Application Volume % of HMDA Activity
Lenders Tested LA Chicago LA Chicago LA Chicago
Black/White 35 49 81,031 78,655 50 47*
Hispanic/Anglo 34 51 80,447 85,214 50 51*
* During the course of testing, one institution in our sample merged with another large lender. HMDA application volume for this lender is difficult to determine and is not represented in these totals.
While the sample was drawn based on lending institution, tests were conducted by visiting individual branch offices, which were randomly selected from all of the institution’s local offices. To select a branch office of a lender, a list of that lender’s local branches was compiled.
Urban institute staff made calls to local offices to verify addresses and determine which branches potential borrowers could visit to receive information on mortgage loans.7 Once the list of local branches was reduced to those providing mortgage information, the branch to be tested was selected randomly from the list. This process was conducted for each lending institution in the sample.
6 Separate samples were drawn for black/white tests and Hispanic/Anglo tests in each site. The medium- to large-volume lenders were likely to be selected for testing in both samples because of their dominance in the local market. However, each sample’s selection of smaller lenders was slightly (and randomly) different.
7 Information on the volume of activity conducted by each branch was not available.
This sampling procedure makes it possible to generalize to the universe of large lending institutions that maintain local offices where first-time homebuyers can inquire in-person about mortgage financing options. It is important to recognize that small lenders, as well as lenders that accept applications only by telephone or over the internet have been excluded from this sample. Their treatment they provide to minority customers (relative to whites) may differ in important ways from the institutions included in this study.
Basic Test Structure
Tests completed during the pilot phase were “paired test” structures involving two individuals, both posing as part of a married household, who were matched on personal, financial, and loan-seeking characteristics so that the primary difference between them was the protected characteristic for which the test was conducted, namely race or national origin. Thus, minority and white testers were paired to form test teams matched by gender and age, and both male and female test teams were used to conduct Hispanic/Anglo and black/white paired tests in Los Angeles and Chicago.
All testers posed as first time homebuyers beginning their search for a single-family home, and called ahead to make an appointment with a loan specialist. During the visits,
testers requested assistance in determining the price range of housing they could afford and the loan amount for which they might qualify. Testers were assigned detailed personal and financial information which they were instructed to provide upon the lender’s request. If testers were provided a recommended home price range and “pre-qualified” for a loan amount, they were told to obtain as much information as possible about any available financing options for which they qualify. Testers were instructed to end their visits by thanking the lender for any assistance provided and to allow the lender to make any arrangements or suggestions for follow-up
contact.
Organizational Structure
Responsibility for the assignment and implementation of tests was shared between the Urban Institute and the local testing organizations contracted in Los Angeles and Chicago. The Urban Institute held a supervisory role, developing all sampling and testing methodologies, training local housing groups and testers, reviewing completed tests, and analyzing testing data.
In addition, UI staff was responsible for sampling institutions and branches, and assigning protected class and financial characteristics to tests. The local housing groups in each site were responsible for the day-to-day implementation of testing. Teams of three managers, two testing coordinators and one quality control supervisor, headed up the testing effort in each site.
Test coordinators hired testers, scheduled tests, assigned personal profiles, and briefed and debriefed testers.
One important lesson from the pre-test experience was that implementing a large number of lender tests and maintaining a consistent level of quality across all the tests required considerable supervision at the local level. The complexity and quantity of information obtained from lender tests required additional quality control beyond that which could be provided by test coordinators. Someone at the local level was needed to provide oversight and a quality review of all test materials immediately following the completion of each. Both local testing
organizations were funded and required to hire a quality control supervisor who could perform this function. Beyond providing the quality review function and additional supervision of the testing process, the quality control supervisors in each site also entered data into a central data base as tests were completed. This local review and data entry process strengthened the quality and consistency of testing―ensuring that testing protocols were followed and verifying the accuracy of data on tester report forms.
Test Authorizations
The authorization to conduct a test was forwarded by the Urban Institute to test coordinators in Los Angeles and Chicago via a Test Site Authorization Form. This form specified the parameters of each test, including the institution and branch to be tested, whether the testers should be white and black or Anglo and Hispanic, the timing and sequence of the test visits, a test identifying number, and any special instructions. Prior to forwarding a Test Site Authorization Form, UI staff made advance phone calls to the institutions and branches selected through the sampling process. Given the fluctuations of the lending market, these calls were necessary to confirm the existence of the branch, make sure the branch was accessible to first-time mortgage seekers, and verify basic contact information. The correct name, address and phone number for the test sites, along with office hours (if known), were provided to the test coordinators on the Test Site Authorization Form. In order to minimize suspicion and the risk of detection, the white or Anglo tester was always instructed to visit prior to the minority tester. It was the
responsibility of the test coordinator to have testers make their visits within 24-120 hours of each other.8
Tester Assignments – Characteristics and Instructions
In each site, test coordinators recruited and selected a pool of testers for participation in the Homeownership Testing Project. Urban Institute staff reviewed the actual characteristics of
8 During the course of the testing, the time spread was increased in some situations. It became apparent that wider spacing was sometimes necessary to accommodate loan officers who maintain erratic work schedules or to reduce the possibility of detection. While most tests were conducted within the 24-120 day spacing, some tester visits were spaced between four to eight days apart.
these testers to ensure that all of them spoke fluent English and were competent to conduct complex mortgage lending tests. Minority testers were all clearly identifiable as either African American or Hispanic. As discussed earlier, the responsibility for developing testers’ assigned characteristics was shared by the Urban Institute and the local test coordinators.
Personal profiles for the test assignments, including tester names and
addresses, personal family information, employers and occupations, and names of local creditors such as local department stores and banks, were developed by the local test coordinators based on their knowledge of the area. Extensive training was provided to test coordinators on how to assign personal characteristics to testers. Instead of using their own home telephone numbers, testers were assigned voice mail numbers so that follow-up calls could be documented in a systematic manner throughout the testing process. Local test coordinators assigned appropriate occupations, employers and length of time on the job to every adult employed in the tester’s fictional household. No tester was assigned to have less than two years on their current job and minority testers were assigned a slightly greater length of employment than their white counterparts. In addition, local test coordinators assigned testers a response to the question of how they happened to hear about the lender, such as through the phone book, on the internet, or from a sign or advertisement.
The financial characteristics assigned to testers were carefully developed and controlled by Urban Institute staff. On every test, the minority tester was slightly more qualified than the white tester. Income (individual and household), debts (monthly payments and amounts owed), and assets (amounts in checking and savings accounts) were provided to testers for every test. These amounts were assigned to ensure that the qualifying ratios (front- and back-end) were very close for both testers, and well within established norms. In Los Angeles, the financial profiles were based around an amount slightly higher than the median single-family home price, while in Chicago, the financial profiles were based around an amount slightly below the median home price. Financial profiles were calculated backwards from the target home price to determine the income and debt amounts necessary to qualify for that amount of home assuming a 30-year conventional fixed-rate loan, at 8 percent interest, with a 5 percent downpayment (see Annex A for more details on assigned financial characteristics). Six profiles were
developed for each MSA and were randomly assigned on each test. In all of the profiles, the testers were somewhat asset constrained, meaning that their available
downpayment―not their qualifying ratios―limited the loan amount for which they qualify. Income was distributed among wage earners in each household in the same manner for both minority and non-minority testers.
In addition to their detailed financial characteristics, testers were assigned:
• Credit Information. All of the testers were assigned profiles with one or two minor credit blemishes, usually a late payment of some kind. Testers were instructed to offer details about their credit situation only if the subject of their credit standing or a credit check was brought up by the lender during the visit.
• Education Level. Adults in the tester households were assigned comparable education levels so that this information could be provided if requested by a lender.
• Current Rent/Period of Time at Current Residence. All testers posed as current renters. The current rent paid by testers was assigned so that the minority tester was always paying a little more in rent than the white tester. No testers resided at their current residence less than three years and minority testers were assigned to have resided at their current residence longer than their white counterparts.
These personal and financial characteristics as well as detailed instructions were provided to each tester prior to conducting a test. The Urban Institute developed a set of instructions or “script” that accompanied every test assignment. The instructions provided testers with a set of tasks they were expected to accomplish during each test, including how to make the call for an appointment, how to state the request, what questions to ask, and how to end the visit. These instructions were designed to add an additional element of control to the testing process so that consistent observations could be obtained about lender conduct. These detailed instructions were consistent across all tests. Testers were provided with only one test assignment at a time and were required to complete that test and document the experience before receiving another test assignment.
Additional instructions were provided to all HTP testers in the Manual for Fair Lending Testers and in the HTP Tester Training Sessions. Examples of additional instructions include:
• Testers must not bargain or negotiate for better financing terms;
• Testers must not provide a social security number or date of birth, and must never authorize a credit check;
• Testers must not agree to pay for any services during the test visits;
• Testers must not state a preference for a particular type of financing, home price, loan amount or monthly mortgage payment; and
• Testers must say that they are not currently working with any real estate agent if asked by a lender.
Testing Protocols
In order to implement testing in a standardized manner across all tests conducted in Los Angeles and Chicago, a set of testing protocols were developed for the pilot phase. Local test coordinators were required to attend a week-long training session on these protocols and were provided a training manual to use as a reference for testing protocols and procedures.
Test coordinators were required to meet with each tester, in person, prior to a test being conducted. At this briefing, testers received their tester assignment and a set of instructions for the upcoming test. During this initial briefing, the test coordinator was responsible for:
• reviewing the test assignment form with the tester and answering any questions about assigned characteristics, instructions and/or testing procedures;
• providing the tester with the appropriate test forms and materials (a
completed test assignment form, a blank test report form, a sufficient quantity of test narrative forms, a notepad, etest coordinator.);
• helping the tester develop a “cheat sheet,” including household income, indebtedness, and assets information (only the financial details that an actual homebuyer might write down in preparation for meeting with a lender);
• clarifying any special instructions; and
• reviewing the procedures for completing the test forms and arranging to meet for a debriefing following the completion of the test.
Test coordinators used a “Checklist for Briefing Testers” that was developed after the pre-test phase to ensure greater consistency in the quality and content of the briefings. A “List of
Reminders to Testers” was included with every tester assignment packet. This list reviewed test procedures and specific areas of tester conduct where mistakes or deviations from test
assignments were found to occur during the pre-test phase.
The testing protocols that testers were instructed to follow when conducting a test can be summarized in five basic steps.
• Step #1 - Obtaining an Appointment. All HTP testers called to arrange “in person”
visits with lenders. Testers were provided detailed instructions on how to complete such calls and how to avoid protracted conversations with loan officers over the phone.
• Step #2 - Making the Initial Request (During the Visit). When testers arrived for their appointments, their first step was to very clearly state (up to three times, if
necessary) that the purpose of their visit is to obtain some help in figuring out a price
range of housing that they might be able to afford and an estimated loan amount for which they might qualify.
• Step #3 - Exchanging Personal/Financial Information. Testers were to be forthcoming and provide income, debts, assets, credit information and other personal and financial characteristics when requested by a lender. Testers were instructed to be precise when providing their financial information and refer to their
“cheat sheets” if necessary. Under no circumstances, however, were testers to provide a social security number or date of birth or authorize a credit check.
• Step #4 - Recording Information on Financing Options Recommended. Testers were required to take notes and record information provided by the lender such as suggested home price range, an estimated loan amount, and details about any financing options recommended.
• Step #5 - Ending the Visit. Testers were instructed to thank the lender for any assistance and allow the lender to suggest any follow-up contact.
Following every test visit, testers were instructed to complete a Test Report Form and a Test Narrative Form. The Test Report Form records testers’ responses to questions about their test experience and details the information that was provided by the lender (see Annex B). The Test Narrative Form is a detailed, chronological, account of the test experience. Additional narrative forms were completed by testers following any phone contact with a lender. Testers were instructed to complete all forms as soon as possible following contact with a lender.
Testers completed the forms based on their recollection of what occurred during the test, and on their review of notes taken and materials obtained during the test.9
All testers were instructed to contact their test coordinator after each test visit to arrange a convenient time to meet as soon as possible following the completion of a test. The purpose of this meeting or “debriefing” was to:
• collect all of the completed test forms, notes and all other materials obtained by the tester;
• review the completed test forms (including the detailed narrative) to make sure the forms were filled out completely and accurately and that they were signed and dated;
9 Because of the complexity of lender testing, the detailed narratives played a particularly important role in quality control. Specifically, test coordinators, quality control supervisors, and Urban Institute staff all reviewed the narrative reports to ensure that testers adhered to HTP protocols and accurately completed the more structured, closed-ended Test Report Form.
• allow the test coordinator to clarify any issues that were raised by the test reports and permit the tester to bring up any concerns about the test or mention any deviations they may have made from the test assignment or protocols; and
• aid the test coordinator in determining whether or not any immediate follow- up was necessary to complete the test (e.g. return visit, phone call).
Test coordinators used a “Debriefing Checklist” to be sure they reviewed the test materials consistently with each tester. Testers were assigned local voice mail telephone numbers so that the local test coordinators could record all follow-up contact received from lenders. Following the tester training session, testers recorded personalized greetings on the voice mail number assigned to them. Test coordinators monitored all assigned voice mail numbers every three to five days for messages. To ensure the integrity of this data, messages could only be retrieved by use of a passcode, which was only provided to Urban Institute staff and the test coordinators. Messages received on tester voice mail numbers were transcribed by the test coordinators in narrative form, and these documents became part of the completed test file. Depending upon the nature of the follow-up contact, test coordinators and Urban Institute staff determined whether it was appropriate for a tester to respond to the follow-up contact that was initiated by the lender.
Replicating HTP Lender Testing Protocols
The testing approach, tester assignment process, testing protocols and test forms described in this chapter were developed exclusively for use in this research project. Other organizations that conduct testing may use different, though no less effective, approaches to mortgage lending testing. Also, the protocols, procedures, training materials and test forms are not intended to limit or restrict the performance of other types of lender testing or establish a legal standard for testing that is conducted to enforce fair lending laws. Organizations that conduct testing to investigate discriminatory lending practices as part of a fair lending enforcement program should understand that these testing protocols and materials were developed solely for use in this research project.
Pre-application mortgage lending testing, whether for research or enforcement, is a complex, costly, and time-consuming activity. This study demonstrates a testing approach which does yield credible and objective observations about lender conduct during the pre- application stage of the mortgage lending process. Organizations that have developed a strong testing capability and expertise in the area of mortgage lending practices may find useful
applications for some of the information described in this chapter. Many organizations, however, including some that routinely conduct other types of fair housing tests may not
currently possess the internal capacity, staff resources, technical knowledge, training, or testers
required to effectively implement pre-application mortgage lending tests. These organizations are cautioned that careful planning, qualified personnel, additional training and substantial resources may be required to overcome some of the significant implementation challenges associated with the type of pre-application mortgage lending tests conducted here.
